On modal tenses and tensed modals
Brenda Laca
To cite this version:
Brenda Laca. On modal tenses and tensed modals. Ch. Nishida / C. Russi. Building a bridge
between linguistic communities of the Old and the New World. Current research in tense,
aspect, mood and modality, Rodopi, pp.163-198, 2012, Cahiers Chronos. <halshs-00323340v2>
HAL Id: halshs-00323340
https://halshs.archives-ouvertes.fr/halshs-00323340v2
Submitted on 4 May 2010
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
On modal tenses and tensed modals*
Brenda LACA
Université Paris 8 – CNRS UMR 7023-SFL
1. Introduction
The study of the interaction between modality and temporality has always
been informed by the insight that these two dimensions are not independent.
A primary source for this insight is the fact that the interpretation of modal
verbs sometimes appears to be fully determined − or at least severely
restricted - by the associated temporal configuration. Thus, (1a) only admits
an epistemic construal, conveying an inference of the speaker as to the truth
of the prejacent1, whereas (1b) excludes such an epistemic construal. This
difference correlates with the presence of perfect morphology below the
modal in (1a) and with the presence of future morphology above the
semimodal in (1b):
(1)
a. He must have left early.
b. He’ll have to leave early.
As for the exclusion of an epistemic construal for (1b), syntacticians have
traditionally assumed that “epistemic” modals cannot be dominated by past
or future tenses (see for instance Cinque 1999), while acknowledging that
this constraint holds at the interpretive, but not necessarily at the overt
morphosyntactic level (Stowell 2004). The interpretive constraint finds its
semantic rationale in the fact that epistemic construals, which convey that the
truth of the prejacent is compatible with or inferable from the beliefs of an
epistemic agent (normally the speaker), are indexical in nature (Papafragou
2005), so that the time of modal evaluation is necessarily set to the now of
the epistemic agent (Boogaart 2005). Tense on the modal in (1b) sets the time
*I gratefully acknowledge support from the Fédération TUL (FRE 2559 CNRS) for
the research program TEMPTYPAC.
Abbreviations: COND= conditional, FUT= future, IMPF= imperfect/ imperfective
past, PP= past participle, PRES= present, SP= simple (perfective) past. Perfect forms
are rendered by the combination of have and the past participle.
1
I’ll follow von Fintel (2005) and von Fintel & Iatridou (2008) in calling "prejacent"
the proposition embedded under the modal.
2
of modal evaluation in the speaker’s future, and thus precludes an epistemic
construal.
The fully determined epistemic construal of (1a), in its turn, can be
derived from the combined action of a model restricting the interaction of
time with the structure of possibilities and a felicity condition on eligible
modal bases. The model of historical necessity captures the intuitive
asymmetry between a fixed past and an open future by ensuring that facts
that precede or are simultaneous with a time t have no alternatives from the
point of view of t, though they may have had from the point of view of a
previous time t’ with regard to which they were future (Condoravdi 2001,
Copley 2002, Kaufmann 2005, Kaufmann, Condoravdi & Harizanov 2006,
Werner 2003, see also Steedman 1997 and Thomason 1984). That is to say,
if we are now in a p-world on account of past or present facts that have
settled the issue, no ¬p-worlds are metaphysically acessible from now;
because they are settled, past and present facts have no metaphysical
alternatives. The diversity condition on the interpretation of modal
statements requires that modals applied to a prejacent p be interpreted against
the background of modal bases that are diverse as to p, i.e. that contain both p
and ¬p-worlds (Condoravdi 2001, Werner 2003)2.
The temporal configuration in (1a) presents the issue of John having
left early or not as settled at the time from which the modal base is accessed
(which in this example coincides with the time of utterance). This being so,
accessible metaphysical modal bases are non-diverse (they contain either
only worlds in which he left early, or only worlds in which he did not) and do
not qualify as a felicitous background for interpretation. However, by far not
every settled fact also happens to be known by an epistemic agent. From the
point of view of an agent who does not know which way things went, an
epistemic modal base contains the alternatives required by the diversity
condition. The epistemic uncertainty construal of (1a) is thus a consequence
of the fact that in such a temporal configuration, only epistemic modal bases
fulfill the diversity condition.
2
The idea that modal statements involving possibility are subject to such a felicity
condition has been around for some time. In fact, Condoravdi (2001: 25) –unlike
Werner (2003)- restricts it explicitly to possibility modals. It is obvious that for this
condition to be extended to necessity modals, modal quantification has to be further
restricted to the “best worlds” as determined by an ordering source (since directly
applying universal quantification to a diverse modal base would result in a
contradiction). The diversity condition thus only makes sense in a framework that
distinguishes between modal bases and ordering sources (see Kratzer 1981,
Kaufmann, Condoravdi & Harizanov 2006). Let me further suggest that the diversity
condition is highly reminiscent of the ban against singleton sets as restrictors of
quantification, which is well attested in other realms of grammar (#Every mother of
John’s is kind, #When Peter dies, there’s sometimes a commotion).
3
As convincingly shown by Werner (2003), the generalisation
according to which modal statements about non-future facts are interpreted
against the background of epistemic or subjective uncertainty holds for a
large number of environments (see also Kaufmann 2005). It thus seems to
capture a central feature of natural modal reasoning.
However, so-called “subjunctive” or “past” modals in English
apparently give rise to an ambiguity when they appear in contexts like (1a).
As shown by its possible continuations, (2) differs from (1a) in being not
only compatible with a construal of epistemic uncertainty (3a), but also with
a counterfactual construal, conveying the speaker’s belief in the falsity of the
prejacent (3b):
(2)
(3)
He should have left early.
a. Let’s check if he in fact did.
b. But he didn’t.
In her thorough dissection of the temporal interpretation of English modals,
Condoravdi (2001) proposes an analysis of the counterfactual construal for
(2) in which the temporal configuration is affected by the additional
morphology on the modal, albeit in a quite indirect way. In this analysis,
“subjunctive” modals allow for a scope reversal mechanism I will label
“covert perfect raising”, by which the perfect operator takes scope over the
modal. This mechanism has the effect of backward-shifting the time from
which the modal base is accessed, thus widening the domain of quantification
for the modal to include metaphysical alternatives that, at the time of
utterance, may have been discarded by the course of events. Widening the
domain triggers the implicature that the speaker does not believe p to be a
current possibility at the time of utterance. The two different construals of (2)
are thus claimed to be associated with a genuine ambiguity in the temporal
configuration involved:
(4)
a. MOD > PERF [epistemic construal]
b. PERF > MOD [counterfactual construal]
My aim in this paper is to explore the interactions between temporal
configuration and modal interpretation by looking at two Romance
languages, French and Spanish. French and Spanish modals exhibit wider
and more transparent morphological options than English modals, and their
patterns of interpretation diverge in significant ways. This description will
show that mere backward-shifting of the time of the modal is not sufficient
for conveying a counterfactuality implicature, but that additional aspectual
ingredients are required, and will call into question the hypothesis that the
4
two different construals of (2) involve a genuine ambiguity, temporal or
otherwise3.
The paper is organised as follows. In section 2, I will clarify the
notion of temporal configuration in a modal environment by resorting to the
distinction between temporal perspective and temporal orientation drawn in
Condoravdi (2001). I will dwell at some length on the way in which temporal
orientation is determined by the aspect of the described situation, in an effort
to bring out its pervasive effects. I will also suggest that, over and above the
time of modal evaluation and the time of the event, we need to take a third
time into account, which I will call “time of decidedness”4, when assessing
statements involving scheduled or foreclosed events. Section 3 is devoted to
French modals and past morphology, and shows that backward-shifting of the
time of the modal does not produce counterfactual construals in this
language. It also provides clear evidence for a scope inversion mechanism by
which tense-aspect morphology on a modal may originate and be interpreted
below the modal, affecting the prejacent itself. Section 4 examines modals
and conditional morphology in French and Spanish. The patterns of
interpretation for sequences combining a modal verb in the conditional with
perfect morphology, though differing in both languages, are entirely
unexpected under Condoravdi’s account. Section 5 is devoted to Spanish
modals and past morphology. Spanish differs quite spectacularly from French
in allowing counterfactual construals for past modals, but these construals
require either perfective aspect on the modal, or perfect morphology on the
infinitive, or both. I will take this fact as an indication that backward-shifting
of the time of the modal only conveys counterfactuality if the issue is taken
to be decided at utterance time UTT-T. Section 6 summarizes the results.
2. Temporal perspective, temporal orientation and decidedness
Temporal configurations in modal environments involve at least two times:
the time from which the modal background is accessed and the time of which
3
Demirdache & Uribe-Etxebarría (2006, 2008) propose an analysis for the English
modals that has the clear advantage of motivating the "covert perfect raising”
mechanism postulated by Condoravdi. This analysis rests on the assumptions (a) that
MOD-T is an interval that cannot be closed, (b) that syntactic movement can
intervene to “rescue” a nonsensical temporal configuration, (c) that temporal
arguments are syntactic objects that can undergo movement.
4
"Decidedness" is the term proposed by Kaufmann (2005) in order to avoid the
potential confusion that may arise from using "settledness" in two different senses,
namely settledness of p, which is equivalent to truth of p for past, present, and predetermined future facts, and settledness of the issue whether p or ¬p. I will try to
stick to the convention of referring to the latter as "decidedness" throughout the text.
5
the temporal property expressed in the prejacent is to hold. As stated above,
differences in the location of the former can have direct consequences on the
domain of metaphysical possibilities, since possibilities decrease with the
course of events (what is settled at a time t, and has therefore no
metaphysical alternatives from the point of view of t, may have had such
alternatives at a previous time t’, when the issue was not yet decided).
Differences in the location of the latter also have direct consequences on the
domain of metaphysical possibilities, on account of the settledness of past
and present facts. The notion of temporal perspective is designed to locate
the time from which the modal background is accessed (henceforth MOD-T),
whereas temporal orientation takes care of the time at which the temporal
property (the described eventuality) is instantiated (henceforth EV-T).
In extensional contexts, the anchor for the temporal perspective can
be safely assumed to be the time of utterance (henceforth UTT-T). In
intensional contexts involving an attitude-bearer, the anchor for the temporal
perspective is the now of the attitude-bearer. This assumption will prove
particularly relevant for the interpretation of modals bearing past morphology
in (free) indirect speech or reported thought contexts. Temporal perspective
(the ordering of MOD-T with regard to its anchor) is determined by tense. In
Condoravdi’s proposal, modals can only appear in the scope of present or
zero tense. Their temporal perspective is thus uniformly simultaneous to their
anchor (“present”), unless the backward-shifting mechanism of “covert
perfect raising” (with “subjunctive” modals for the past) ensures a “past”
temporal perspective. Thus, both (5a) and (5b) can only have a “present”
temporal perspective, (5a) because of the lack of perfect morphology, (5b)
because of the lack of the required “subjunctive” morphology enabling
“covert perfect raising”. Only (5c) allows for an interpretation (the
counterfactual construal) for which a “past” temporal perspective is assumed:
(5)
a. He should leave early.
b. He must have left early.
c. He should have left early.
As for temporal orientation, a recurrent observation in the literature
is that the aspectual class of the described situation determines temporal
orientation. In English, eventive predicates uniformly receive a forwardshifted, “future” orientation (6a), whereas stative predicates receive by
default a simultaneous, “present” orientation (6b) and are compatible with a
“future” orientation in the presence of future-oriented adverbials (6c):
(6)
a. John may get drunk/ talk with the Dean.
b. John may be drunk.
6
c. John may be drunk by the time we arrive.
The progressives of eventive predicates behave like states, with a default
“present” orientation (7a) that can be overriden by a future-oriented adverbial
(7b):
(7)
a. John may be getting drunk/ talking with the Dean.
b. John may be getting drunk/ talking with the Dean by the time we arrive.
In Condoravdi’s framework, perfect infinitives determine a “past”
orientation. Note, however, that if one adopts the view that perfect infinitives
contribute the post-state of an event (see Kamp & Reyle 1993 among many
others), the facts are entirely parallel to those for states and progressives,
with a default “present” orientation for the post-state (8a) that can be
overriden by a future-oriented adverbial in a “future perfect” configuration
(8b):
(8)
a. John may have gotten drunk/ talked to the Dean.
b. John may have gotten drunk/ talked to the Dean by the time we arrive.
Condoravdi’s analysis accounts for the obligatory “future”
orientation of eventives and the “present” or “future” orientation of states by
assuming that modals uniformly expand the time of evaluation forward.
MOD-T is modelled as an open interval [t, _) stretching indefinitely into the
future of t. The initial boundary of MOD-T is set by the temporal
perspective. In examples (6-8) this initial boundary can, for the reasons
explained above, only be present (simultaneous with UTT-T). The open
interval is required to include EV-T in the case of eventive predicates (so that
the first moment of EV-T cannot precede the first moment of MOD-T), but
only to overlap with it in the case of stative predicates (thus allowing, but not
forcing, a reading in which the first moment of EV-T precedes the first
moment of MOD-T). The perfect, in Condoravdi’s analysis, contributes an
anteriority relation which takes the initial boundary of MOD-T as anchor and
thus ensures a “past” orientation.
The pattern of temporal orientation according to which eventives
determine a “future” orientation and states a “present” one is pervasive in
English. As shown by Iatridou (2000) and Kaufmann (2005), it also holds for
the present and “fake past” (“subjunctive”) antecedents of conditionals,
which constitute modal environments and are standardly analyzed as
containing a covert modal operator. However, the same pattern can also be
recognised in simple present sentences, in which no obvious element in the
context introduces modality, and even in infinitival complements of attitude
7
verbs requiring simultaneous readings (the class labelled B-verbs in Abusch
2004). In fact, whereas states in the simple present or as infinitival
complements of B-verbs have a “present” orientation (9a-b), eventives have a
“future” one −- unless they can be interpreted generically/habitually 5.
Future-oriented eventives give rise in such contexts to so-called scheduled
interpretations (Copley 2008, Kaufmann 2005, Kaufmann, Condoravdi &
Harizanov 2006). As a consequence, only events that are amenable to
scheduling can naturally occur therein, as shown by the differences in
acceptability between (10a-b) and (11a-b):
(9)
(10)
(11)
a. John is sick.
b. John is said to be sick.
a. ??John gets well.
b. ??John is said to get well.
a. Our train leaves from platform C.
b. Our train is said to leave from platform C.
The hypothesis that MOD-T is a forward-expanding open interval
[t, _) is posited as a means of capturing the influence of aspectual class on
temporal orientation. However, it fails to account for temporal orientation in
two crucial points. Firstly, the future orientation of eventives also holds
outside modal environments. The “futurate” interpretation of eventive
infinitives embedded under B-verbs provides clear evidence that this property
is independent of the embedding context. B-verbs actually require
simultaneity of the infinitival complement with the reported attitude (Abusch
2004), but the time of instantiation of an eventive infinitive is no less
forward-shifted in these contexts than it is in the context of an embedding
modal. Secondly, as recently observed by Copley (2008), the assumption that
states are required to overlap with MOD-T does not explain why states do
not shift forward freely, but are normally interpreted as “present” in the
absence of a future anchor they could be simultaneous to. In fact, the futureoriented adverbial in example (6c), as well as those in (7b) and (8b), are
essential for the future-orientation of a state.
As emphasized by Copley (2008), the correlation of aspectual class
with temporal orientation takes the form in (12a-b), and is thus actually
stronger than predicted in Condoravdi’s analysis:
(12)
5
a. EVENTIVES → “FUTURE” ORIENTATION
b. STATES → “PRESENT” ORIENTATION
Following Smith (1991), I assume that generic/habitual interpretations constitute
derived stative situations. Their temporal orientation is thus predictably "present".
8
Conceiving of MOD-T as a forward expanding open interval [t, _)
helps capture the fact that temporal orientation is NON-PAST in modal
environments, but it does not contribute much to explain the correlation in
(12). The crucial question concerning temporal orientation is why states are
simultaneous to the initial boundary of MOD-T (the time from which the
modal base is accessed) if nothing else intervenes, whereas eventives cannot
be simultaneous, but only follow it.
Phenomena of aspectual-class driven temporal location involving a
difference between states and eventives have been extensively explored both
in discourse (see for instance Kamp & Reyle 1993) and in the study of
tenseless languages (see Smith 2008). The general simultaneity requirement
for states can be derived from a crucial property of their temporal structure:
they are totally homogeneous situations, lacking endpoints. Since their initial
or final transitions are not part of their temporal structure (Smith 1991), they
are not available for temporal location. Furthermore, because of their
temporal homogeneity, states can be evaluated at points in time (Dowty
1979). As a result, states will have a default “present” orientation. Eventives,
by contrast, are not totally homogeneous and their temporal structure has
endpoints. Since they involve change, they cannot be evaluated at points in
time (Dowty 1979). This is the source of the Bounded Event Constraint,
which excludes a “present” orientation for eventives (Smith 2008). In a
NON-PAST environment, such as is contributed by a modal verb, but also
by the present tense in (11a) or by the attitude verb in (11b), the only
available possibility for eventives is thus “future” orientation.
In English, the Bounded Event Constraint holds for all nonaspectualised eventives, unless they can be understood as habituals/generics,
i.e. as derived states. Eventives in the progressive (cf. (7a-b) above) behave
as states. Whether the progressive is interpreted as changing the temporal
structure of the eventuality by providing a derived state, or as the expression
of a viewpoint aspect excluding the initial and final endpoints of the
eventuality, while entailing their initial transition (Smith 1991), the resulting
structure is the same: a lack of endpoints that will result in a default “present”
orientation. The same reasoning can be extended to perfect infinitives (cf. 8a
–b) above), if they are held to contribute the post-state of an eventuality: the
post-state will have a default “present” orientation, and the described
eventuality will, as a consequence, strictly precede the anchor.
The pattern in (12a-b) is also detectable in Spanish and French, but
with one significant difference. The temporal orientation of states is, as in
English, by default “present”, and the temporal orientation of achievements is
uniformly “future”. But it seems possible, and much easier in French than in
Spanish, to get a “present” orientation for atelic eventive predicates, such as
activities (and those accomplishments that might be reinterpreted as
activities). This is illustrated in (13a-b) and (14a-b) by the contrast between
9
‘destroy the evidence’, an accomplishment that is not easily coerced into an
activity, and the activity predicate ‘talk to the Dean’:
(13) French
a. Il doit
détruire les preuves.
He MUST.PRES destroy the evidence
‘He must destroy the evidence’
b. Il doit
s’entretenir avec le Doyen.
He MUST.PRES talk to the Dean
‘He must talk to the Dean’/ ‘He must be talking to the Dean’
(14) Spanish
a. Debe
destruir las pruebas.
MUST.PRES.3.SG destroy the evidence
‘S/he must destroy the evidence’
b. Debe
conversar con el Decano.
MUST.PRES.3.SG talk to the Dean
‘S/he must talk to the Dean’/ ??‘S/he must be talking to the Dean’
As indicated by the glosses, the telic predicates in (13a) and (14a) only give
rise to a future-oriented interpretation. By contrast, the atelic predicate in
(13b) can also give rise in French to a present-oriented interpretation, thus
conveying an inference of the speaker as to a settled present fact (epistemic
uncertainty reading). This interpretation is much less natural in Spanish. This
peculiarity in the possible temporal orientation of atelic eventives extends to
the other contexts mentioned above, most notably to the present and “fake
past” (“subjunctive”) antecedents of conditionals, to simple present
sentences, and to infinitival complements of B-verbs. The pattern observed
for French and Spanish is summarized in (15):
(15)
a. TELIC EVENTIVES → “FUTURE” ORIENTATION
b. STATES→ “PRESENT” ORIENTATION
c. ATELIC EVENTIVES → “FUTURE” ORIENTATION/ “PRESENT” ORIENTATION
[FRENCH]
d. ATELIC EVENTIVES → “FUTURE” ORIENTATION/ ??”PRESENT” ORIENTATION
[SPANISH]
The different patterns exhibited by English, Spanish, and French
clearly correlate with the availability and “obligatorification” of a progressive
form and with its effects on simple (non-progressive) forms. English has a
fully grammaticalised progressive and therefore does not normally allow
ongoing readings for the simple forms of eventive predicates. The
progressive in Spanish is more widely available than it is in French (for the
dubious status of être en train de + Infinitive in French, see Smith 1991, for a
comparison between French and Spanish, see Bertinetto 2000). Though not
10
entirely excluded, ongoing readings for non-progressive forms are difficult to
obtain in Spanish (see Laca 2005a). The question of the “obligatorification”
of a progressive form – as a function of its polarization effects on simple
(non-progressive) forms –- is notoriously complex, and cannot be explored in
this paper. But it is important to note that the patterns of temporal orientation
in (12) and (15) correspond to quite general patterns of .aspectual-class
driven temporal location for forms lacking explicit temporal or aspectual
morphology, which are known to be sensitive to the availability and
obligatoriness of contrasting forms (see Smith 2008). What seems to be at
work in temporal orientation is a strategy for ordering the temporal trace of
an eventuality relatively to a point in time when strict anteriority is not an
option and no aspectual morphology further specifies the relation6. For the
complement of modal verbs, this anchoring point is the initial boundary of
MOD-T.
Recall now that temporal perspective, as determined by tense, orders
the initial boundary of MOD-T with regard to UTT-T, and that temporal
orientation, as determined by aspectual class, is the location of EV-T with
regard to the initial boundary of MOD-T. Thus, there is no direct relation
linking EV-T with UTT-T. The question that arises at this point is whether
temporal perspective and temporal orientation capture all the relevant aspects
of temporal configurations in modal environments. The assessment of modal
statements involving scheduled or foreclosed events suggests that this is not
the case.
As discussed in the introduction, modal statements about decided
issues are interpreted against the background of epistemic or subjective
uncertainty as a way of fulfilling the diversity requirement for the modal
base. If an issue is not yet decided, objective uncertainty obtains and ensures
diversity for a metaphysical or circumstantial modal base. Of course,
objective uncertainty invariably gives rise to subjective uncertainty (for a
formalisation of this fact, see the lack-of-foreknowledge principle in
Kaufmann, Condoravdi & Harizanov 2006). But the readings that are
normally recognised as “epistemic” readings for modals, those that convey
that the truth of the prejacent is compatible with or inferrable from the beliefs
of a speaker who does not know for sure, are those involving epistemic
uncertainty about decided facts. Decidedness of the issue (objective
6
In the solution proposed by Demirdache & Uribe-Etxebarria (2006, 2008), the
voidness of the aspectual head which should determine the relationship between Ev-T
and Ast-T (the interval for which the assertion is made, which corresponds to Smith’s
1991 “interval of visibility”) plays a crucial role in temporal orientation. I cannot
discuss their solution here, but I believe that the basic intuition is the same as the one
suggested in the text, namely that the problem of temporal orientation is a possible
ramification of the much larger problem of “unspecified” aspect.
11
certainty) is a necessary condition for epistemic readings. The problem is
that, as discussed in Condoravdi (2001), we sometimes obtain “epistemic”
readings in connection with future orientation. The example proposed by
Condoravdi is revealing:
(16)
He will meet with one senior administrator. It has been decided who he will
meet but I don’t know who it is. He may see the dean. He may see the
provost.
Epistemic readings for modals embedding future oriented eventives involve
scheduled or pre-determined events (Copley 2008). Although EV-T is in the
future, there are present relevant facts (the existence of a plan determining
who he is supposed to see in (16) above) that are held to entirely determine at
UTT-T the truth-value of the corresponding “futurate” prejacent. Such cases
are thus instances of deterministic or “check the facts”-futures, as opposed to
indeterministic or “wait-and-see” futures (see Bonomi & Del Prete 2008).
Temporal perspective and temporal orientation do not exhaust all the relevant
aspects of this temporal configuration. Crucially, we need to capture the fact
that the prejacent, notwithstanding the future orientation of the eventive
predicate, has a knowable truth value at UTT-T. I propose to capture this fact
by assuming a relationship directly linking the “time of decidedness” to
UTT-T. Schematically, the configuration corresponding to He may see the
dean in the above context can be represented as in (17):
(17)
UTT-T
MOD-T
EV-T
t-decidedness
A second motivation for the necessity of dissociating a “time of
decidedness” from temporal orientation is provided by temporal mismatches
giving rise to counterfactual readings, as exemplified in (18):
(18)
Laure might have won tomorrow’s race.
Such examples present the eventuality described in the prejacent as
foreclosed by the intervention of some event precluding its occurrence. The
temporal adverbial clearly locates EV-T in the future of UTT-T. In
Condoravdi’s analysis, the mechanism of covert perfect raising would
determine a past temporal perspective, and the eventive infinitive (not
affected by the perfect) a future temporal orientation. As far as I can see, this
temporal configuration fails to capture a non-cancellable feature of (18),
12
namely that, at UTT-T, the issue of Laure’s winning or not tomorrow’s race
is already decided. At a more general level, genuine counterfactuality
presupposes decidedness at UTT-T, and it is not easy to see how the
combination of a past temporal perspective with a future orientation is apt to
convey decidedness. As we will see in the next sections, a past temporal
perspective is clearly not sufficient for conveying counterfactuality, and the
analysis of counterfactual readings in terms of covert perfect raising poses
serious problems in French and Spanish alike. This will lead us to abandon
the idea that the contribution of perfect morphology to counterfactuality
consists in setting back the time from which the modal base is accessed. On
the contrary, I will propose that perfect morphology directly links the “time
of decidedness” to UTT-T 7.
3. French modals and past morphology
By contrast with English modals, French modals can be fully and
transparently inflected and thus bear past morphology. Past morphology on a
modal does not always determine a past temporal perspective, but even in the
cases where it does, it never warrants a counterfactual construal, thus calling
into question the assumption that past temporal perspective is the factor that
triggers counterfactual construals.
Let us first review the possible interpretations of sequences
containing modals in the Passé composé, which we will treat, with some
degree of simplification, as an expression for a perfective past. Such
sequences admit two interpretations: (a) an entailing interpretation, in which
there is a clear inference that the eventuality took place in the actual world,
and (b) an epistemic interpretation, in which the past occurrence of the
eventuality is inferred or conjectured against the background of subjective
epistemic uncertainty. The sequences exclude a continuation in which nonoccurrence of the eventuality is asserted, i.e. they are not compatible with
counterfactuality:
(19)
7
a. Pierre a pu
prendre le train de 3.50, [#mais il ne l’a pas fait].
Pierre has CAN.PP take the train of 3.50, [#but he didn’t]
#‘Pierre managed to take the 3.50 train, but he didn’t’
#‘Pierre may have taken the 3.50 train, but he didn’t’
b. Pierre a dû
payer une amende importante, [#mais il ne l’a pas
Locating the 'time of decidedness' before UTT-T is crucial in the case of some
configurations giving rise to non-cancellable counterfactual implicatures, such as
temporally mismatched sentences with future adverbials (see above) and statives in
the perfect with present oriented adverbials, as in He should/could have been in his
office by now.
13
fait].
Pierre has MUST.PP pay a fine important, [#but he didn’t]
#’Pierre was forced to pay a huge fine, but he didn’t’
#’Pierre must have paid a huge fine, but he didn’t’
I will not dwell further on the entailing interpretation, which has been
extensively analyzed by Hacquard (2006). I agree in principle with her idea
that such interpretations involve a lower modal under the scope of tense and
aspect (see also Borgonovo & Cummins 2007)8.
If morphology is taken at face value, the epistemic interpretation of
(19a-b) poses two related problems. Firstly, it clashes with the assumption
that telic eventive infinitives, since they give rise to future orientation
according to the pattern in (15a), cannot convey the decidedness required by
epistemic interpretations in the absence of explicit contextual clues.
Secondly, it apparently violates the constraint according to which, in
epistemic interpretations conveying inferences and conjectures, the time of
modal evaluation coincides with the now of the epistemic agent. The fact is,
however, that in such cases overt morphology does not seem to be a reliable
guide to interpretation.
There is a widespread consensus that these configurations involve
scope inversion between the modal and tense-aspect morphology (Stowell
2004, Borgonovo & Cummins 2007, Demirdache & Uribe-Etxeberria 2006,
2008, Hacquard 2006). As suggested by the glosses, the epistemic
interpretation of (19a-b) does not exhibit a past temporal perspective with a
future orientation, but rather a present temporal perspective with a past
orientation. For all practical purposes, (19a-b) are equivalent to the more
transparent linearisations in (20a-b):
(20)
a. Pierre peut avoir pris le train de 3.50.
Pierre CAN.PRES have take.PP the train of 3.50
‘Pierre may have taken the 3.50 train’
b. Pierre doit avoir payé une amende importante.
8
Let me observe, however, that entailing interpretations are always related to the
intentions or desires of an agent, which may either overcome or be thwarted by
opposing forces. It is not clear to me how Hacquard's analysis can capture this fact. In
any case, the entailing interpretation seems to vanish whenever the described
eventuality is not of the kind that can be intentionally brought about or prevented by
an agent, so that (i) and (ii) only have the epistemic interpretation:
(i)
Il a pu
pleuvoir.
It has CAN.PP rain ‘It may have rained’
(ii)
Pierre a dû
trébucher.
Pierre has MUST.PP stumble ‘Pierre must have stumbled’
14
Pierre MUST.PRES have pay.PP a fine important
’Pierre must have paid a huge fine’
As pointed out by Stowell (2004), possible syntactic accounts for
scope inversion all come at a certain cost. The less costly solution appears to
be a reconstruction-like analysis in which tense-aspect morphology originates
between the modal and the verbal projection and raises to combine with the
modal, while being interpreted in its original position, as schematically
represented in (21)9:
(21)
Modal
[T [VP] ]
__ /
But even though being less costly, reconstruction-like analyses should be
motivated by stronger evidence than simple semantic intuition. As far as I
can see, the strongest empirical support for this analysis comes from an
observation formulated originally by Tasmowski (1980)10: the choice of past
morphology appearing on the modal in an epistemic interpretation matches
exactly the choice that would be mandatory for the main verb of the prejacent
in the absence of the modal. Such mandatory choices for a Passé composé,
for an Imparfait and for a Plus-que-parfait are illustrated below. They are
determined by the aspectual class of the described situation and by the
temporal adverbial in (22a-b), and additionally by overall temporal coherence
in a sequence like (23a):
(22)
a. Marie a écrit/
*écrivait ce roman en moins d’un an.
Marie has write.PP/ *write.IMPF this novel in less of a year
‘Marie wrote this novel in less than a year’
b. Marie détestait/ * a détesté Pierre depuis longtemps.
Marie hate.IMPF/ *has hate.PP Pierre since long time
‘Marie had hated Pierre for a long time’
(23)
a. Pierre partit de Londres un jeudi .
Il
y
était
arrivé
la veille.
9
See however Demirdache & Uribe-Etxeberria (2008) for an alternative solution,
proposed for Spanish, in terms of lowering of a temporal head expressing anteriority
to a void aspectual head below the modal. This movement is semantically motivated
by the need to avoid an uninterpretable temporal configuration. I cannot dwell on the
details of their analysis, but I suspect that configurations with pluperfect morphology
above the modal, as in (26a) below, are not amenable to this solution in their
framework, since we have a second anteriority relation that must be expressed
somewhere.
10
See Borgonovo & Cummins (2008) for additional evidence.
15
Pierre partir.SP of London a Thursday. He there be.IMPF arrive.PP the daybefore
‘Pierre left London on a Thursday. He had arrived there the day before’
b. #Pierre partit de Londres un jeudi.
Il y
est
arrivé la veille.
Pierre partir.SP of London a Thursday. He there be.PRES arrive.PP the daybefore
Il y
arrivait
la veille
He there arrive.IMPF the day-before
#‘Pierre left London on a Thursday. He has arrived/was arriving there the
day before’
The crucial observation is that a past-tense devoir only gets an epistemic
(inferential) interpretation if the past morphology replicates this choice:
(24)
(25)
(26)
a. Marie a dû
écrire ce roman en moins d’un an. √EPISTEMIC
Marie has MUST.PP write this novel in less of a year
‘Marie must have written this novel in less than a year’
b. Marie devait
écrire ce roman en moins d’un an. *EPISTEMIC
Marie MUST.IMPF write this novel in less of a year
‘Marie was supposed/ had to write this novel in less than a year’
a. Marie devait
détester Pierre depuis longtemps.
Marie MUST.IMPF hate Pierre since long time
‘Marie must have hated Pierre for a long time’
b. *Marie a dû
détester Pierre depuis longtemps.
Marie has MUST.PP hate Pierre since long time
a. Pierre partit de Londres un jeudi . Il avait
dû
y arriver.
P. leave.SP of London a Thursday. He have.IMPF MUST.PP there arrive.PP
la veille
the day-before
‘Pierre left London on a Thursday. He must have arrived there the day
before’
b. #Pierre
partit de Londres un jeudi . Il a
dû/
P.
leave.SP of London a Thursday. He have.PRES MUST.PP /
devait
y
arriver la veille
MUST.IMPF there arrive the day-before
On the basis of this evidence, I will adopt the consensual view that
past morphology in the epistemic readings of (19a-b) does not determine
temporal perspective but reflects temporal orientation. Though realized on
the modal, past morphology only has scope over the prejacent. This view
immediately circumvents the difficulties associated with an epistemic
construal (a) whose time of modal evaluation does not coincide with the now
of the epistemic agent and (b) which involves eventive infinitives normally
determining a future orientation. It should be stressed that adopting this view
forces us to assume that there is “inner tense”, i.e. tense below the modal, in
epistemic readings conveying inferences or conjectures (see Steedman 1997).
The central piece of the analysis is overt movement without interpretive
16
consequences, a sort of “overt tense-aspect raising” that contrasts with the
“covert perfect raising” assumed by Condoravdi (2001) for the counterfactual
construal of English modals for the past.
Let me recapitulate the findings up to now. A Passé composé never
gives rise to a counterfactual construal. If it originates and is interpreted
above the modal, we obtain an entailing interpretation. If it originates and is
interpreted below the modal, it conveys a (present) inference or conjecture
with regard to a past event. Imparfait and Plus-que-parfait on a modal can
also originate and be interpreted below the modal and convey (present)
inferences or conjectures with regard to past events.
However, unlike the Passé composé, the Imparfait and the Plus-queparfait on a modal are also compatible with epistemic readings in a different
temporal configuration, in which they also fail to determine a past temporal
perspective. The Imparfait is known to exhibit both genuine past tense uses
and anaphoric or “zero-tense” uses in past intensional contexts. In the latter
case, it conveys simultaneity with a past act of thinking or speaking. I will
refer to the former as “free” imperfects and to the latter as “bound”
imperfects. “Bound” imperfects give rise to epistemic readings of modals
bearing past-morphology, as shown in (27a-b). These readings are licensed
by the past in the matrix, and disappear in the context of an embedding
present tense (28a-b):
(27)
(28)
a. Le
voyant arriver si tôt,
j’ai pensé qu’il devait
avoir pris le
him seeing arrive so early, I thought that he MUST.IMPF have take.PP the
premier train.
first train
‘Seeing him arrive so early, I thought that he must have taken the first train’
b. Tout suggérait que la lettre pouvait
avoir été envoyée par un proche.
all suggested that the letter CAN.IMPF have been sent
by a friend
‘Everything suggested that the letter might have been sent by a friend’
a. ?? Le voyant arriver si tôt, je pense qu’il devait
avoir pris le
him seeing arrive so early, I think that he MUST.IMPF have take.PP the
premier train.
first train
b. ?? Tout suggère que la lettre pouvait avoir été envoyée par un proche.
All suggests that the letter CAN.IMPF have been sent by a friend
The mechanism of overt and semantically inert “tense raising” we have just
discussed can result in equivalent Plus-que-parfait versions for (27a-b). In
fact, (29a-b) and (27a-b) alternate without semantic difference, in the same
way in which (19a-b) and (20a-b) above alternate:
(29)
a. Le voyant arriver si tôt,
j’ai pensé qu’il
avait
dû
prendre
him seeing arrive so early, I thought that he have.IMPF MUST.PP take
17
le premier train
the first train
‘Seeing him arrive so early, I thought that he must have taken the first train’
b. Tout suggérait que la lettre avait
pu
être envoyée par un proche.
all suggested that the letter have.IMPF CAN.PP been sent by a friend
‘Everything suggested that the letter might have been sent by a friend’
Epistemic readings with “bound” Imparfaits, which may appear
massively in free indirect speech contexts, do not really challenge the
constraint according to which epistemic construals are tied to the now of the
epistemic agent and thus require a present temporal perspective. In (free)
indirect speech or reported thought contexts, the Imparfait is an anaphoric
tense, interpreted simultaneously to the now of the attitude bearer: the
temporal perspective in these contexts is that of a “present of the past”11.
This leaves us only “free” Imparfaits as possible candidates for
determining a past temporal perspective. The question to be answered at this
point is whether a “free” Imparfait determining a past temporal perspective,
in combination with an eventive infinitive imposing a future temporal
orientation, gives rise to a counterfactual construal (recall that in
Condoravdi’s analysis, the counterfactual construal is triggered by the
combination of a past temporal perspective and a future temporal
orientation). Hacquard (2006: 77) gives an affirmative answer to this
question, asserting that (30) “is compatible with Jane actually taking the train
and going to London, but the preferred reading is that she in fact didn’t”:
(30)
Jane pouvait
prendre le train pour aller à Londres.
Jane CAN.IMPF take the train
for go to London
‘Jane could go to London by train’
I must disagree with her on this point. I believe that there is a fundamental
difference between the counterfactual construal discussed by Condoravdi and
the putatively preferred interpretation of (30). A genuine counterfactual
construal should be at the very least incompatible with the certainty of the
speaker as to the truth of the proposition being evaluated. This is a property
that “subjunctive” modals for the past in English have (cf. 31a), but that the
configuration in (30) admittedly lacks:
11
Boogaart (2005), though rejecting the “zero-tense” hypothesis, has shown the
pervasive presence of epistemic modals bearing past morphology in such contexts. He
emphasizes the fact that such epistemic readings require simultaneity of MOD-T
with a past act of thinking or speaking that must be overtly expressed or otherwise be
clearly inferrable from the context.
18
(31)
a. Jane could have gone to London by train, #so she did.
b. Jane pouvait prendre le train pour aller à Londres, et c’est ce qu’elle a
fait.
Simple compatibility with – or even preference for- a continuation
asserting the falsity of the prejacent is not enough for attributing
counterfactuality to an environment. Note, furthermore, that the French
configuration in (30) – a good candidate for past temporal perspective and
future orientation – differs from the English configuration in (31a) in another
important respect. The former is infelicitous in contexts in which the falsity
of the prejacent is established, the latter is not:
(32)
a.# Le patient est mort, bien qu’ il pouvait être sauvé.
The patient has died, although he CAN.IMPF be saved
#‘The patient died, although he could be saved’
b. The patient died, although he could have been saved.
On the basis of the compatibility of the Imparfait in (30) with an assertion as
to the truth of the prejacent, and of its incompatibility in (32a) with contexts
in which its falsity is established, we conclude that ”free” Imparfaits no more
license counterfactual construals in French than the Passé composé does. The
indicative contexts reviewed up to this point show that the simple
combination of past temporal perspective with future orientation is not an
available strategy for conveying counterfactuality in French. As we will see
in the next section, counterfactual readings in French require special
morphology on the modal.
4. Modals and conditional morphology in French and Spanish
Conditional morphology (Imparfait morphology on a future stem) plays a
central role in the expression of counterfactuality, since it typically appears in
the consequent of counterfactual or “future-less-vivid” conditionals (Iatridou
2000). If we grant that Romance modals bearing conditional morphology are
the nearest equivalent to English “subjunctive” modals (von Fintel & Iatridou
2008), the pattern in (33a-b) appears to indirectly confirm Condoravdi’s
analysis of English. As shown by the possible continuation as to the falsity of
the prejacent, a counterfactual construal is available when the perfect appears
above the modal (33a), but it seems excluded when the perfect appears below
the modal (33b):
(33)
a. Marie aurait
pu
gagner la course (mais elle l’a perdue).
M. have.COND CAN.PP win the race (but she lost).
19
‘M. might/ could have won the race’
b. Marie pourrait avoir gagné la course (#mais elle l’a perdue).
M.
CAN.COND have win.PP the race (#but she lost).
‘M. possibly/ allegedly/ reportedly won the race’
Note that this pattern, while apparently providing evidence for the link
between the [PERFCOND> MODAL] configuration and the counterfactual
construal, confirms at the same time that additional morphology on the modal
(conditional morphology in French, a “subjunctive” modal in English) is a
necessary ingredient for this construal. This fact is not accounted for in
Condoravdi’s analysis.
Furthemore, upon closer inspection the pattern linking the position
of perfect morphology above the modal to counterfactuality, and its position
below the modal to epistemic uncertainty, turns out to be less conclusive, and
to partially differ for possibility and necessity modals.
In fact, the configuration [PERFCOND>MODAL] in (33a) is not only
compatible with counterfactuality, but also with a construal of epistemic
uncertainty. This is shown not only by the acceptability of a continuation
conveying epistemic uncertainty – pour autant que nous sachions “for all we
know” is unobjectionable both after (33a) and after (33b) – but also in
naturally occurring examples as (34a), in which the [PERFCOND>MODAL]
configuration with the possibility modal is used for establishing a hypothesis,
not for discarding it. (34b) shows that devoir in the [PERFCOND>MODAL]
configuration is also compatible with a construal of epistemic uncertainty:
(34)
a. Au vu du terrain,
un hélicoptère aurait
bien pu
venir de
At sight of-the terrain, a helicopter
have.COND well CAN.PP arrive from
derrière la colline proche sans être entendu.
behind the hill nearby without being heard
‘Due to the characteristics of the terrain, a helicopter might well have
arrived from behind the nearby hill without being heard’
b. Pierre aurait
dû
arriver chez lui il y
a une heure.
P.
have.COND MUST.PP arrive at him it there has an hour
Appelle-le pour t’assurer qu’il est bien là.
‘Pierre should have arrived home an hour ago. Call him up to check if he’s
in fact there’
Differences between the possibility and the necessity modal arise for
the configuration in which the modal appears above the perfect. In
[MODALCOND>PERF] configurations, devoir – but not pouvoir – is accepted
(though dispreferred) with continuations conveying counterfactuality (35a),
and occurs naturally (though infrequently) in contexts where no epistemic
20
uncertainty is involved, but rather the falsity of the prejacent is asserted or
otherwise established (35b):
(35)
a. Marie devrait
avoir déjà
mis son adversaire hors-jeu.
M.
MUST.COND have already put.PP her rival
off-game
Je ne comprends pas pourquoi elle traîne.
‘Marie should have already defeated her rival. I don’t understand what
she’s waiting for’
b. L’ ossification est très en retard et sur les radiographies, on note la
persistance des
cartilages de conjugaison
à un âge où
ils devraient
avoir disparu.
to an age where they MUST.COND have disappear.PP
‘Ossification lags considerably behind and on the X-rays, we see articular
cartilages subsisting at an age when they should have disappeared’
The data are extraordinarily complex, but they clearly call into question the
assumption of a contrast between a construal of epistemic uncertainty and a
counterfactual construal that correlates with a difference in both temporal
perspective and temporal orientation. The ambiguity between both construals
resurfaces at least in part in French, a language that has the syntactic patterns
required to resolve it, and the patterns of ambiguity are sensitive to the
quantificational force of the modal. As we will presently see, the Spanish
facts, though diverging from what we have just seen for French, reinforce
these misgivings.
The first surprising observation is that, although in Spanish, as in
French, perfect morphology can be realised either above or below a modal
bearing conditional morphology, the pattern of interpretation for the latter
case ([MOD.COND>PERF]) is parallel to the English, and not to the French
pattern. Crucially, sequences of this type are compatible with counterfactual
continuations, and are also perfectly natural in contexts in which the falsity of
the prejacent is established. We saw above that this was only marginally
possible for devoir in French (see 35a-b above). In Spanish, by contrast, this
is the preferred option both for possibility and for necessity modals:
(36)
a. María podría
haberse quedado en México,
María CAN.COND have-REFL remain.PP in Mexico
pero prefirió
volver a su país.
but prefer.SP.3SG return to her country
‘María could have stayed in Mexico, but she preferred to go back home’
b. Le dio
a los banqueros, cuando debería
haber apoyado a los
him give.SP to the bankers, when
MUST.COND have support.PP to the
deudores.
debtors
S/he gave to the bankers, when s/he should have supported the debtors’
21
c. En vez de quedarse callado, tendría
que haber intervenido.
in time of keep-REFL quiet, HAVE.COND that have intervene.PP
‘Instead of keeping quiet, s/he should have intervened’
The configuration [MODCOND>PERF] is also compatible with a construal of
epistemic uncertainty, as in French or, for that matter, English:
(37)
a. El accidente podría
haberse
originado en un
the accident CAN.COND
have-REFL
originate.PP in a
fallo del
motor.
failure of+the engine
‘The accident might have been caused by engine failure’
b. María ya
debería
haber llegado a su casa.
María already MUST.COND have arrive.PP to her house
¿Porqué no llamas para confirmar?
‘Maria should have arrived home by now. Why don’t you call up and
check?’
c. Tendría
que haber intervenido, pero no sé si lo hizo.
HAVE.COND.3.SG that have intervene.PP
‘S/he should have intervened, but I don’t know if s/he did’
Spanish patterns like French, and unlike English, in admitting constructions
in which perfect morphology is realised above the modal. However,
[PERFCOND>MODAL] configurations are on the whole less frequent than
[MODCOND>PERF] configurations, and this difference in frequency is much
more pronounced in the case of necessity modals12. [PERFCOND> MODAL]
configurations are fully compatible with counterfactual construals (i.e. the
patterns in (36a-c) can be subject to “overt perfect raising” without
discernable meaning differences), but epistemic construals seem to be only
possible, and marginally so, with the possibility modal. Example (38)
parallels the French example (34a) above:
(38)
Por ello, estos científicos piensan que la primera célula habría
by this, these scientists think
that the first
cell have.COND
podido viajar protegida dentro de un meteorito desde el planeta en que
12
A search of the Davies corpus (http://www.corpusdelespanol.org) for 20th
century Spanish gives the following results:
poder 'can'
deber 'must'
tener que 'have to'
MOD.COND haber V-PP
139
33
22
haber.COND MOD-PP V-INF
63
0
4
22
CAN.PP travel protected inside of a meteorite from the planet in which
se originó.
REFL originate.SP
‘For this reason, these scientists believe that the first cell might have
travelled inside a meteorite from the planet where it originated’
Before proceeding, let us compare our results for French and
Spanish. In spite of the complexity of the data, a relatively clear pattern
emerges from this comparison. French and Spanish have a linearisation
option that is compatible both with a counterfactual construal and with
epistemic uncertainty. In French, this option is the configuration in which the
perfect appears above the modal, in Spanish it is the configuration in which
the perfect appears below the modal. Let us call this the “preferred option”
for each language. It is the dispreferred option that appears to lend some
support to the assumption of a link between “perfect raising” and
MODCOND + PERF. INF
PERFCOND MOD + INF
French
Spanish
√EPIST
√EPIST √CF
CF *pouvoir
(√) devoir
√EPIST √CF
EPIST (√) poder √CF
*deber
*tener que
Table 1: Possible construals with conditional modals and perfect
morphology
counterfactuality, but the support is rather weak. In Spanish a higher perfect
precludes epistemic uncertainty with necessity modals, but marginally admits
it with the possibility modal; in French, a lower perfect precludes
counterfactuality with the possibility modal, but marginally admits it with the
necessity modal. The patterns of interpretation are summarized in Table 1
above, in which the preferred option for each language is highlighted.
The patterns of interpretation summarized in Table 1 give rise to a number of
related questions:
(i) What is the contribution of conditional morphology?
(ii) Does the difference in the linearisation of perfect morphology reflect
a difference in temporal configuration?
(iii) Is there a genuine epistemic/counterfactual ambiguity in contexts
involving conditional morphology?
(iv) What accounts for the different behaviour of possibility and
necessity modals?
23
As to the first question, conditional morphology in French and
Spanish is either anaphoric on a past time of thinking or speaking, in its role
as “future of the past”, or it is licensed by a modal environment. Outside
(free) indirect speech or reported thought contexts, conditional morphology
appears on modal verbs, on verbs expressing wishes or preferences, and in
the consequent of (possibly covert) conditional sentences. In such contexts,
conditional morphology acts as a domain-widener: it signals that worlds are
being accessed that lie outside the common ground, i.e. outside the domain of
what is being taken for granted by the speaker13. An indication of the
domain-widening effect of conditional morphology is the fact that modals in
the conditional are much more easily compatible with negative belief
assertions than indicative modals. (39a-b) exemplifies this contrast for the
possibility modal in a context of objective uncertainty in Spanish, (40a-b) for
the necessity modal in a context of subjective uncertainty in French:
(39)
a. ??Puede producirse un accidente, pero no creo que vaya a pasar nada.
CAN.PRES produce.REFL an accident
‘There may be an accident, but I don’t think anything will happen’
b. Podría producirse un accidente, pero no creo que vaya a pasar nada.
CAN.COND produce.REFL an accident
‘There might be an accident, but I don’t think anything will happen’
(40)
a. #Jean doit être chez lui, mais je ne pense pas qu’il y soit.
Jean MUST.PRES be at home
‘Jean must be at home, but I don’t think he is’
b. Jean devrait être chez lui, mais je ne pense pas qu’il y soit.
Jean MUST.COND be at home
‘Jean should be at home, but I don’t think he is’
Note that, in the light of the contribution of conditional morphology
in modal contexts, the motivation for backward-shifting MOD-T in order to
access metaphysical alternatives discarded by the further course of events is
not compelling. If conditional morphology in modal contexts can itself widen
the domain of quantification to sets of worlds that do not belong to the
common ground, there is simply no need for a second domain widener14.
13
On domain-widening by “subjunctive” morphology, see von Fintel (1999) and, for
Spanish, Alonso-Ovalle (2002).
14
Common grounds are modelled as epistemic modal bases (Condoravdi 2001). In the
mechanism proposed by Condoravdi (2001), widening the domain of metaphysical
alternatives by backward-shifting MOD-T signals that the domain of quantification
for the modal is partly outside the common ground. In the present proposal, widening
the domain of quantification of the modal to include epistemic alternatives not
contained in the common ground (the contribution of conditional morphology) may
24
This takes us to our second question: what is the contribution of
perfect morphology? A look at what happens in the absence of perfect
morphology provides the answer. If the situation described in the prejacent is
eventive, and thus determines a “future” temporal orientation, both the
epistemic and the counterfactual construal disappear: we obtain weak modal
readings involving objective uncertainty.
(41)
(42)
French
a. Il pourrait
rencontrer des difficultés.
he CAN.COND meet
of+the difficulties
‘He could meet with some difficulties’
b. Ils devraient
aboutir à un accord.
they MUST.COND reach to an agreement
‘They should reach an agreement’
Spanish
a. Podría
producirse un accidente
CAN.COND produce.REFL an accident
‘There might be an accident’
b. Deberían
llegar a un acuerdo.
MUST.COND.3.PL arrive to an agreement
‘They should reach an agreement’
c. Tendria
que tomar el tren de las 3.50
HAVE.COND.3.SG that take the train of the 3.50
‘S/he ought to take the 3.50 train’
However, if the situation described in the prejacent is stative, and thus
determines a “present” temporal orientation, we obtain both epistemic
uncertainty and counterfactual construals, as shown by the two possible
continuations in the following examples:
(43)
(44)
a. Il pourrait être tranquilement chez lui
he CAN.COND be calmly
at home
(pour autant que je sache/ au lieu d’étre ici)
‘He could be comfortably at home (for all I know/ instead of being here)
b. Ce livre devrait
être sur l’étagère de droite
this book MUST.COND be on the shelf of righthand
(va vérifier/ et non pas là où il est)
‘This book ought to be on the right-hand shelf (go check/ and not where it
is)’
a. Podría
estar tranquilo en su casa
CAN.COND.3.SG be calmly in his house
have the effect of recovering a world history that has been discarded by the course of
events.
25
(por lo que sé/ en lugar de estar aquí)
‘He could be comfortably at home (for all I know/ instead of being here)’
b. Ese libro debería
estar en el estante de la derecha
this book MUST.COND be in the shelf of righthand
(fíjate/y no donde está)
‘This book should be on the right-hand shelf (go check/ and not where it
is)’
c. Ese libro tendría
que estar en el estante de la derecha
this book HAVE.COND that be in the shelf of righthand
(fíjate/y no donde está)
‘This book ought to be on the right-hand shelf (go check/ and not where it
is)’
The only difference between (41a-b/42 a-c) above and the examples (43 a-b/
44 a-c) resides in the fact that there is no presumption of decidedness in the
former case, but that there is such a presumption in the latter. In fact, the
presumption of decidedness is a necessary condition shared by epistemic and
counterfactual construals: both require that the issue be settled at UTT-T. In
the case of the epistemic construal, the issue is settled, but the speaker does
not know in which way it has been settled, in the case of the counterfactual
construal, the issue is settled, and the speaker does know in which way. I
assume that perfect morphology - whatever its relative position in the
preferred option in French or Spanish - contributes this presumption of
decidedness, by setting the moment at which p- and ¬p-world-histories
branch before UTT-T.
In the line of reasoning we are pursuing, the fact that a construction
is felicitous both in contexts of epistemic uncertainty as to p and in contexts
in which the falsity of p is either established or is to be suggested does not
amount to a genuine ambiguity. Conditional modals signal that the domain
has been widened to include alternatives that do not belong to the common
ground. Perfect morphology signals that the the issue is decided at UTT-T.
The difference between the epistemic and the counterfactual
construal hinges on the knowledge attributed to the speaker: counterfactuality
arises under the further assumption that the speaker knows which way things
went. The difference between both construals thus appears as an instance of
the usual jump from “not believing that p is the case” to “believing that p is
not the case”.
The different behavior of necessity and possibility modals in these
configurations is an unexpected result, which requires careful research. I will
only advance two remarks which may indicate the direction this research
should take.
First of all, note that the patterns in Table 1 can be interpreted in the
sense of an affinity of the possibility modal with construals of epistemic
uncertainty: pouvoir cannot be associated with the assumption that the
26
speaker knows which way things went in the dispreferred configuration in
French, and poder can be associated with epistemic uncertainty in the
dispreferred configuration in Spanish.
Secondly, though compatible with a construal of epistemic
uncertainty, the necessity modal in the conditional is not appropriate in some
of the most typical contexts for the epistemic readings of modals. Recall that
such readings typically involve an inference or a conjecture of the speaker.
Now, necessity modals in the conditional cannot convey inferences as to
causes, as shown by the contrasts in (45a-b) and (46a-b):
(45)
(46)
a. El accidente puede/
debe/
podría
haber sido causado
The accident CAN.PRES/ MUST.PRES/ CAN.COND have be.PP caused
por un fallo del motor.
by a failure of+the engine.
‘The accident may/must/might have been caused by engine failure’
b. #El accidente debería
haber sido causado por un fallo del motor.
The accident MUST.COND have be.PP caused by a failure of+the engine
‘The accident should have been caused by engine failure’
a. Il y a des traces sur la neige. Les fugitifs
There are traces on the snow. The runaway
ont pu/
ont dû/
auraient pu
emprunter ce chemin.
have CAN.PP/ have MUST.PP/ have.COND CAN.PP borrow this path.
‘There are footprints on the snow. The runaway may/must/might have taken
this path’
b.# Il y a des traces sur la neige. Les fugitifs
There are traces on the snow. The runaway
devraient
avoir emprunté ce chemin.
MUST.COND have borrow.PP this path
‘There are footprints on the snow. The runaway should have taken this path’
Though compatible with a construal of epistemic uncertainty, necessity
modals in the conditional do not seem to give rise to interpretations in which
the occurrence of the event is inferable from the information available to the
speaker. Rather, they signal that occurrence of the event was expected in the
light of previous facts (if everything went as planned or as determined by a
previous causal chain). If this intuition is confirmed, it could mean that the
preference for epistemic or metaphysical/circumstantial modal bases, rather
than being associated to temporal perspective, is associated to modal force.
To summarize, the facts discussed in this section do not validate the
hypothesis according to which construals of epistemic uncertainty and
counterfactual construals involve a difference in temporal configuration
deriving from the scope of the perfect operator. The domain-widening effect
stems from conditional morphology, and the perfect operator indicates that
the issue has been decided before UTT-T. Rather than with a semantic
27
contrast, the linearisation differences correlate with a preference for “higher
perfects” in French and for “lower perfects” in Spanish. While modals
bearing conditional morphology in French and Spanish do not lend support to
the idea that counterfactual construals are determined by a past temporal
perspective, we will see in the next section that modals bearing past
indicative morphology in Spanish provide such support.
5. Modals and past morphology in Spanish
Modals bearing past morphology in Spanish exhibit all the interpretive
possibilities we have discussed for French in section 3. They give rise to
entailing readings with the perfective past (47a-b), and to epistemic readings
in which the morphology on the modal reflects the morphology that would
appear in the prejacent (cf. 48a-b and 49a-b):
(47)
(48)
(49)
a. Pedro pudo tomar el tren de las 3.50
Pedro CAN.SP take the train of the 3.50
‘Pedro managed to take the 3.50 train’
’Pedro might have taken the 3.50 train’
b. Pedro debió
pagar una multa importante.
Pedro MUST.SP pay a fine important
‘Pedro was forced to pay a huge fine’/’Pedro must have paid a huge fine’
a. En aquella época, el correo quedaba / *quedó lejos.
‘In those times, the post office was.IMPF/ *was.SP far away’
b. En aquella época, el correo debía quedar / *debió quedar lejos
‘In those times, the post office MUST.IMPF/ *MUST.SP be far away’
a. El que lo mató
*era
/ fue el mayordomo.
‘The one who killed him *was.IMPF/ was.SP the butler’
b. El que lo mató
*debía /
debió ser el mayordomo.
‘The one who killed him *MUST.IMPF/ MUST.SP the butler’
Together with the possibility of epistemic interpretations for (47a-b), the
contrasts of acceptability in examples (48b) and (49b) indicate that the same
mechanism of “overt tense raising” without interpretive consequences we had
assumed for French is also attested in Spanish. A minor, but relevant
difference between both languages is that in the case of perfect or pluperfect
morphology, Spanish prefers the linearisations with a lower perfect in (50a-b)
to the “perfect raised” linearisations in (51a-b). This confirms that Spanish
shows a preference for lower, while French shows a preference for higher
perfects, as was the case for conditional modals:
(50)
a. Puede
haberse
producido un accidente.
CAN.PRES.3.SG. have-REFL produce.PP an accident
28
(51)
‘There may have been an accident’
b. Debía
haber llovido mucho, porque las calles estaban mojadas.
MUST. IMPF.3.SG. have rain.PP a lot, because the streets were.IMPF wet
‘It must have rained a lot, because the streets were wet’
a. ??Ha
podido producirse un accidente.
have.PRES.3.SG. CAN.PP produce- REFL an accident
‘There may have been an accident’
b. ?? Había
debido llover mucho, porque las calles estaban mojadas.
have.IMPF.3.SG MUST.PP rain a lot, because the streets were.IMPF wet
‘It must have rained a lot, because the streets were wet’
In any case, the really significant difference is that in Spanish, in contrast
with French, past morphology on a modal can give rise to a counterfactual
construal. However, counterfactuality only arises in the presence of
perfective past morphology on the modal (52), of perfect morphology on the
infinitive (53), or of a combination of both (54).
(52)
(53)
(54)
María pudo /
debió
escaparse (pero no lo hizo)
María CAN.SP / MUST.SP escape-REFL
‘María could / should have fleed (but she didn’t)’
María podía /
debía/
tenía
que haberse escapado
María CAN.IMPF / MUST.IMPF/ HAVE.IMPF +that have-REFL escapePP
(pero no lo hizo)
‘María could / should have fleed (but she didn’t)’
María pudo /
debió/
tuvo
que haberse escapado (pero no lo hizo)
María CAN.SP / MUST.SP/HAVE.SP+ that have- REFL escapePP
‘María could / should have fleed (but she didn’t)’
The possibility of counterfactual construals for (52-54) clearly supports the
hypothesis that backward-shifting the time from which the modal base is
accessed is a possible domain-widening strategy for conveying
counterfactuality. But at the same time, it shows that past temporal
perspective and future orientation are not sufficient for the counterfactuality
effect: a modal in the imperfective past embedding an eventive infinitive is
no more apt to convey counterfactuality in Spanish than it is in French (see
the discussion of examples (30-32) above). The question arises at this point
as to the necessary ingredient that perfective aspect and/or perfect
morphology contribute to the counterfactual construal. I would like to
suggest that this is, again, decidedness. Perfective aspect and/or perfect
morphology on the infinitive are signaling that p- and ¬p-world-histories
have branched before UTT-T.
29
MOD.SP + INF
MOD.IMPF + PERF. INF
MOD.SP + PERF. INF
TEMPORAL PERSPECTIVE
READING
past
present (“tense raising”)
past
‘present of the past’
present (“tense raising”)
past
present (“tense raising”)
past
entailing
epistemic
counterfactual
epistemic
epistemic
counterfactual
epistemic
counterfactual
Table 2: Temporal perspective and the readings of past modals
The above sequences with past modals show multiple ambiguities.
Thus, in the absence of the counterfactual continuation, (52) can have an
entailing, a counterfactual and an epistemic reading (Borgonovo & Cummins
2007). Next to its counterfactual reading, (53) can also have en epistemic
reading, provided that the imperfect is a “bound” imperfect in a (free)
indirect speech or reported thought context (see the discussion of examples
(27a-b) above), or that the context licenses a pluperfect prejacent (see the
discussion of example (26a) and example (50b) above). Finally, (54) can
have both a counterfactual and an epistemic reading. Note that the perfect
infinitive in (54), which has been held to be redundant (Bosque 1999),
eliminates on the one hand the entailing reading and, on the other hand, is
necessary for the counterfactual construal of the possessive modal tener
‘have’. These ambiguities correlate with genuine differences in temporal
perspective, which are summarized in Table 2 above.
Borgonovo (2008) has recently proposed an analysis in which the
three readings for simple past modals in Spanish correlate with different
scope configurations among three heads: the modal (MOD), the past tense
(PAST) and perfect or perfective aspect (PERF). The three possible scope
configurations are given in (55a-c):
(55)
a. PAST > PERF > MOD: entailing reading
b. PAST > MOD > PERF: counterfactual reading
c. MOD > PAST > PERF: epistemic reading
This analysis accounts for the difference between entailing and
counterfactual readings, which share a past temporal perspective but differ in
the fact that the first embeds a bare infinitive, whereas the second embeds an
aspectualised one. It also explains why the perfect infinitive eliminates
entailing readings: a perfect infinitive unambiguously signals the presence of
aspect below the modal. The configuration attributed to epistemic readings
30
corresponds to the source configuration for the mechanism of overt and
semantically inert “tense raising” that we have assumed both for Spanish and
for French. Finally, the assumption that the Spanish simple past is a complex
morphological category, contributing both a tense and an aspect head,
squares well with the hypothesis I have defended elsewhere (see Laca
2005a), according to which the Romance simple past is the only simple tense
carrying an aspectual specification. I think this analysis is on the right track,
although it opens a number of questions as to why and how morphosyntactic
configurations involving modals give rise to such peculiar scope
permutations15.
As to the differences between Spanish, which admits counterfactual
readings for past modals under the conditions expressed in Table 2, and
French, which doesn’t, Borgonovo (2008) suggests that they correlate (a)
with the fact that (contemporary informal) French lacks a simple past tense, a
suggestion that has also been advanced by Demirdache & Uribe-Etxeberria
(2008), and (b) with the fact that Spanish, unlike French, can use the simple
infinitive instead of the perfect infinitive, as shown in (56a-b):
(56)
a. après avoir mangé/ *après manger
after have eaten / after eat
b. después de haber comido/ después de comer
after
of have eaten / after
of eat
‘after having eaten/ after eating’
Since French apparently must overtly express perfect on an infinitive, on the
one hand, and on the other hand excludes sequences with two perfects, as
shown in (57), it lacks the conditions for expressing configuration (55b),
which is the source of counterfactual readings with perfective pasts:
(57)
*Elle a pu
avoir gagné.
she has CAN.PP have win.PP
However, these correlations do not explain why modals in the Imparfait
in combination with perfect infinitives do not give rise to counterfactual
readings in French. Recall that a sentence like (58) can only have an
epistemic reading:
15
Note that peculiar scope effects with modals seem to be pervasive. They not only
surface in connection with tense and aspect, but also in connection with other
operators, as for instance other modals and negation in the can't seem-construction
(cf. Jacobson 2006) or subject quantifiers giving rise to "epistemic containment" (von
Fintel & Iatridou 2003).
31
(58)
Pierre devait/
pouvait
avoir pris le premier train.
Pierre MUST.IMPF/CAN.IMPF have take.PP the first train
I surmise that this impossibility correlates with several other differences in
the possible range of interpretations for the imperfective past in French and
Spanish which cannot be explored in this paper.
6. Concluding remarks
In this paper, we have examined the interactions between tense-aspect
morphology and the interpretation of modals in French and Spanish in the
light of the analysis proposed by Condoravdi (2001) for English. Although
we have found evidence both for scope inversion phenomena and for the
hypothesis that a past temporal perspective may contribute to counterfactual
interpretations, no clear evidence was found for the idea that “perfect raising”
contributes to backward-shifting MOD-T and thus to the domain-widening
that results in counterfactual readings. Both French and Spanish allow for the
surface order PERF>MOD, but such configurations only give rise to
counterfactual readings in the presence of an extra element of modal-tense
morphology, the conditional, which is itself a domain-widener. The preferred
linearisation options in such cases, which differ for both languages, are
felicitous both in contexts of epistemic uncertainty and in counterfactual
contexts, an entirely unexpected fact that calls into question the existence of a
genuine epistemic-counterfactual ambiguity with conditional modals –which,
we assume, roughly correspond to English “subjunctive” modals. The
different behaviour of possibility and necessity modals seems to indicate that
necessity
modals
in
the
conditional
always
rely
on
metaphysical/circumstantial modal bases.
The inability of modals in the imperfective past or in the conditional
to trigger counterfactual readings when they embed simple eventive
infinitives, together with the possibility of counterfactuality with stative
infinitives or in the presence of perfect morphology (modulo some relevant
differences between both languages) indicates that a past temporal
perspective or domain-widening morphology are not sufficient for
counterfactuality. We have hypothesized that the role of perfect morphology
is to provide the missing element of decidedness for eventives, by indicating
that the relevant branching between p and ¬p world histories has taken place
before UTT-T. For statives, decidedness is ensured by the fact that, in the
absence of other temporal clues, they will be interpreted as simultaneous
with UTT-T.
We have remained agnostic as to the usefulness of conceiving
MOD-T as an open interval stretching indefinitely into the future.
32
Phenomena of aspectually-driven temporal orientation arise with the same
characteristics in contexts not involving modals, and if they are actually
governed by the Bounded Event Constraint, they involve temporal location
with regard to a point (the initial boundary of MOD-T). On the other hand,
we have shown that –if temporal orientation is conceived as the relation
between EV-T and MOD-T- we need a third relation directly linking the
“time of decidedness” of the prejacent to UTT-T. This is precisely the
relation expressed by perfect morphology in the context of conditional
modals and in the counterfactual readings of past tense modals in Spanish.
A number of questions relating to compositionality remain open for
further research. They relate mainly to the mechanism we have labelled
“tense raising”, which allows for material originating and being interpreted in
the prejacent to be realised on the modal, and to the possibility exhibited by
simple infinitives embedded under perfective past modals in Spanish to be
interpreted as perfect infinitives.
33
References
Abusch, D. (2004). On the temporal composition of infinitives, in J. Guéron ;
J. Lecarme, (eds), 27-54.
Alonso-Ovalle, L. (2002). Spanish DE-clauses are not always in the right
mood, in J. Quer et al., (eds), Romance Languages and Linguistic
Theory 2001, Amsterdam : John Benjamins.
Bertinetto, P.-M. (2000). The progressive in Romance, as compared with
English, in O.Dahl, (ed.), Tense and Aspect in the Languages of Europe,
Berlin : Mouton/de Gruyter, 559-604.
Bonomi, A. ; Del Prete, F. (2008). Evaluating future-tense sentences in
changing contexts, Ms. University of Milan.
Boogaart, R. (2005). The past and perfect of epistemic modals, submitted to
Cahiers Chronos 2005.
Borgonovo, C. (2008). Modales ambiguos, Ms, Université Paris 8/Université
de Laval.
Borgonovo, C. & Cummins, S. (2007). Tensed Modals, in J. Eguren ; O.
Fernández Soriano, (eds), Coreference, Modality, and Focus,
Amsterdam : John Benjamins, 1-18.
Bosque, I. (1999). ¿Qué sabe el que sabe hacer algo? Saber entre los
predicados modales, in K. Korta ; F. García Murga, (eds), Palabras.
Victor Sánchez de Zavala in memoriam. Filosofía 3 (EHU), 303-323.
Cinque, G. (1999). Adverbs and Functional Heads. A Cross Linguistic
Perspective, Oxford : Oxford University Press.
Condoravdi, C. (2001). Temporal interpretations of modals. Modals for the
present and for the past, in D. Beaver et al. (eds), The Construction of
Meaning, Stanford : CSLI Publications., 59-88.
Copley, B. (2002). The Semantics of the Future, Ph.D. dissertation, MIT.
Copley, B. (2008). Temporal orientation in conditionals, in J. Guéron ; J.
Lecarme (eds), 59-78.
Demirdache, H. ; Uribe-Etxeberria, M. (2006). Scope and anaphora with time
arguments: the case of ‘perfect modals’, to appear in H. Harley ; R.
Folli, (eds), Lingua 118.
Demirdache, H. ; Uribe-Etxeberria, M. (2008). Morfosintaxis e interpretación
de los verbos modales, in A. Carrasco Gutiérrez, (ed.), Tiempos
compuestos y formas verbales complejas, Madrid-Frankfurt :
Iberoamericana,
Dowty, D. (1979). Word Meaning and Montague Grammar, Dordrecht :
Kluwer.
Hacquard, V. (2006) Aspects of Modality. Ph.D. dissertation, MIT.
von Fintel, K. (1999). The presupposition of subjunctive conditionals, in
U.Sauerland ; O.Percus, (eds), The Interpretive Tract, MIT Working
Papers in Linguistics, 29–44.
34
von Fintel, K. (2005). Modality and Language, to appear in Encyclopaedia of
Philosophy, 2nd edition.
von Fintel, K. ; Iatridou, S. (2003), Epistemic containment, Linguistic Inquiry
34, 2 : 173-198.
von Fintel, K. ; Iatridou, S. (2008). How to say ought in foreign: the
composition of weak necessity modals, in J. Guéron ; J. Lecarme, (eds),
115-142.
Guéron, J. ; Lecarme, J., (eds), (2004). The Syntax of Time, Cambridge,
Mass.:
MIT Press.
Guéron, J. ; Lecarme, J., (eds), (2008). Time and Modality, Berlin : Kluwer.
Iatridou, S. (2000). The grammatical ingredients of counterfactuality,
Linguistic Inquiry 31, 2 : 231-270.
Kamp, H. ; Reyle, U. (1993). From Discourse to Logic, Dordrecht : Kluwer.
Jacobson, P. (2006), I can’t seem to figure this out, in B. Birner ; G. Ward,
(eds), Drawing the Boundaries of Meaning: Neo-Gricean Studies in
Semantics and Pragmatics in Honor of Laurence R. Horn, Amsterdam :
John Benjamins, 157-175.
Kaufmann, S. (2005). Conditional truth and future reference, Journal of
Semantics 22 : 231-280.
Kaufmann, S. ; Condoravdi, C. ; V. Harizanov, V. (2006). Formal approaches
to modality, in W. Frawley, (ed.), The Expression of Modality, Berlin :
Mouton/de Gruyter, 71-106.
Kratzer, A. (1981). The notional category of modality, in H.-J. Eikmeyer ; H.
Rieser, (eds), Worlds, Words, and Contexts, Berlin : de Gruyter, 38-74.
Laca, B. (2005a) Périphrases aspectuelles et temps grammatical dans les
langues romanes, in H. Bat-Zeev Schyldkrot ; N. Le Querler, (éds), Les
périphrases verbales, Amsterdam : John Benjamins, 47-66.
Laca, B. (2005b). Tiempo, aspecto y la interpretación de los verbos modales
en español, Lingüística (ALFAL) 17 : 9-43.
Papafragou, A. (2005). Epistemic modality and truth conditions, in A.
Klinge ; Müller, (eds), Perspectives on Modality, Amsterdam : John
Benjamins. 1688-1702.
Smith, C. (1991). The Parameter of Aspect, Dordrecht : Kluwer.
Smith, C. (2004). The domain of tense, in J. Guéron ; J. Lecarme, (eds), 597620.
Smith, C. (2008) Time with and without tense, in J. Guéron ; J. Lecarme,
(eds), 227-249.
Steedman, M. (1997). Temporality, in J. v. Benthem ; A. ter Meulen, (eds),
Handbook of Logic and Language, London : Elsevier, 895-939.
Stowell, T. (2004). Tense and Modals, in J. Guéron ; J. Lecarme, (eds), 621636.
35
Tasmowski, L. (1980). Un devoir épistémique. Travaux de linguistique 7 :
43-58.
Thomason, R. (1984). Combinations of tense and modality, in D. Gabbay ; F.
Guenthner, (eds), Handbook of Philosophical Logic: Extensions of
classical logic, Dordrecht : Reidel, 135-165.
Werner, Th. (2003). Deducing the Future and Distinguishing the Past:
Temporal Interpretation in Modal Sentences in English. PhD
Dissertation, Rutgers University.